Research map / Vision-language and multimodal / Human and embodied modeling
Density of States Prediction of Crystalline Materials via Prompt-guided Multi-Modal Transformer
Namkyeong Lee, Heewoong Noh, Sungwon Kim, Dongmin Hyun, Gyoung S. Na, Chanyoung Park
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add energy embeddings to E3NN | E3NN · Phonon DOS | MSE: 0.210 -> 0.200 | Improved |
| add energy embeddings to Graph Network | Graph Network · Electron DOS | MSE: 0.264 -> 0.240 | Improved |
| add energy embeddings to MLP | MLP · Phonon DOS | MSE: 0.309 -> 0.251 | Improved |
| add energy embeddings to E3NN | E3NN · Electron DOS | MSE: 0.296 -> 0.291 | No clear effect |
| add energy embeddings to E3NN | E3NN · Band gap from predicted DOS | MSE: 0.896 -> 1.605 | Hurt |
| add energy embeddings to Graph Network in OOD atom species | Graph Network · OOD # atom species | MSE: 0.460 -> 0.455 | No clear effect |
| add energy embeddings to Graph Network in OOD crystal system | Graph Network · OOD crystal system | R2: 0.505 -> 0.534 | Improved |
| DOSTransformer vs best baseline with energy (Electron DOS) | DOSTransformer · Electron DOS | MSE: 0.240 -> 0.221 | Improved |
| fine-tune only prompts instead of all parameters (ablation) | DOSTransformer · OOD crystal system fine-tuning | MSE: 0.365 -> 0.355 | Improved |
| remove either global or system loss (ablation) | DOSTransformer · in-distribution | MSE | Hurt |
| inject crystal system as one-hot into atom features instead of prompt before self-attention (ablation) | DOSTransformer · Electron DOS | MSE: 0.221 -> 0.222 | Hurt |
| inject prompt into atom features instead of before self-attention (ablation) | DOSTransformer · Electron DOS | MSE: 0.221 -> 0.227 | Hurt |
Closest papers by meaning
- Crystalformer: Infinitely Connected Attention for Periodic Structure Encoding (ICLR 2024 · similarity 0.86)
- Complete and Efficient Graph Transformers for Crystal Material Property Prediction (ICLR 2024 · similarity 0.85)
- Vector Field Oriented Diffusion Model for Crystal Material Generation (AAAI 2024 · similarity 0.84)
- Wyckoff Transformer: Generation of Symmetric Crystals (ICML 2025 · similarity 0.84)
- A Diffusion-Based Pre-training Framework for Crystal Property Prediction (AAAI 2024 · similarity 0.84)
- Learning the Electronic Hamiltonian of Large Atomic Structures (ICML 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
